Modified Newtonian Dynamics, an Introductory Review

Riccardo Scarpa
European Southern Observatory, Chile
E-mail rscarpa@eso.org
Abstract. By the time, in 1937, the Swiss astronomer Zwicky measured the velocity dispersion of the Coma cluster of
galaxies, astronomers somehow got acquainted with the idea that the universe is filled by some kind of dark matter. After
almost a century of investigations, we have learned two things about dark matter, (i) it has to be non-baryonic -- that is,
made of something new that interact with normal matter only by gravitation-- and, (ii) that its effects are observed in
stellar systems when and only when their internal acceleration of gravity falls below a fix value a0=1.210-8 cm s-2. Being
completely decoupled dark and normal matter can mix in any ratio to form the objects we see in the universe, and indeed
observations show the relative content of dark matter to vary dramatically from object to object. This is in open contrast
with point (ii). In fact, there is no reason why normal and dark matter should conspire to mix in just the right way for the
mass discrepancy to appear always below a fixed acceleration. This systematic, more than anything else, tells us we
might be facing a failure of the law of gravity in the weak field limit rather then the effects of dark matter. Thus, in an
attempt to avoid the need for dark matter many modifications of the law of gravity have been proposed in the past
decades. The most successful and the only one that survived observational tests -- is the Modified Newtonian
Dynamics. MOND posits a breakdown of Newton's law of gravity (or inertia) below a0, after which the dependence with
distance became linear. Despite many attempts, MOND resisted stubbornly to be falsified as an alternative to dark matter
and succeeds in explaining the properties of an impressively large number of objects without invoking the presence of
non-baryonic dark matter. This suggests MOND is telling us something important about gravity in the weak field limit.
In this paper, I will review the basics of MOND and its ability to explain observations without the need of dark matter.

INTRODUCTION
As far back as 1983, after realizing that mass discrepancies are observed in stellar systems when and only when
the acceleration of gravity falls below a fix value[1], M. Milgrom proposed a modification of Newtonian dynamics
(MOND) as an alternative to non-baryonic dark matter[2,3,4]. Milgrom's proposal states that gravity (or inertia) does
not follow the prediction of Newtonian dynamics for acceleration smaller than a0=1.210-8 cm s-2, as derived from
galaxy rotation curves fit[5]. Below this acceleration the behavior of gravity changes becoming asymptotically

a = aN a0 , where aN is the usual Newtonian acceleration. The transition from Newtonian to MOND regime
occurs for accelerations that remain undetected within the solar system where the strong field of the sun is
dominating all dynamical processes. To be more precise, we see that even the acceleration by Mercury on Pluto is
above a0, so that the there is no way in the solar system to test the validity of Newtonian dynamics down to the
acceleration regimes typical of galaxies. As a consequence, we have no direct proof of its validity below a0, and its
applicability in the realm of galaxies is not guaranteed.
Since the seminal papers by Milgrom[2,3,4] a few authors have worked on this subject showing how this simple -though revolutionary -- idea explains many properties of galaxies without the need of non-baryonic dark matter (
e.g., [5,6,7,8]). MOND is so successful that, as a minimum, it is telling us the exact functional form of the force in
galaxies. Any theory of galaxy and structure formation must therefore be able to reproduce the MOND
phenomenology. MOND is also effective in describing the dynamics of galaxy groups and clusters[9,10] and, with
some approximations, gravitational lensing[11,12]. For a recent review see [13].
As I shall show, MOND has been successfully tested in a very large number of cases and, were it not for dark
matter being thoroughly ingrained in the mind of astronomers, by now there should be little doubt about its
relevance in describing the universe. Nowadays, the problem is no longer to show whether MOND works -- because
this was already done -- rather that MOND and dark matter give conceptually different but operationally equivalent

description of cosmic phenomena. Because of this, astronomers have decided to stick to dark matter and continue to
add more and more arbitrary entities, once called epicycles, to save the appearances.
To make progress in this field one has to find a way to show MOND - or equivalently dark matter -- effects in
places where no one is expecting them. In the first place MOND should be tested in the laboratory (but see below
Milgrom's opinion about this). Also, one can study the dynamics of very wide double stars, globular clusters or other
small structure where the effects of dark matter are believed to be negligible. If this will be done --successfully of
course -- in a sufficiently large number of cases, it will be no longer possible to claim that dark matter is the
explanation and new physics will be required. However, before moving too far ahead, lets start comparing MOND
predictions to real data.

MOND BASICS
The MOND acceleration of gravity a is related to the Newtonian acceleration aN by

aN = a( a /a0 ).

The constant a0=1.210-8 cm s-2 is meant to be a new constant of physics.

The interpolation function (a/a0) admits the asymptotic behavior =1 for a>>a0, so to retrieve the Newtonian
expression in the strong field regime, and =a/a0 for a<<a0.
With these assumptions, the
! weak acceleration limit of gravity is:

a = aN a0 =

GMa0
,
r

(1)

with dependence 1/r on distance r from the body of mass M generating the field.
Different though functionally equivalent expression for have been used in the literature (e.g., [3,7,14]).
Currently the most used function is:

( a /a0 ) =

a /a0
1+ (a /a0 ) 2

which works well in describing galactic rotation curves and has the advantage of being simple. Solving for a the
MOND acceleration of gravity reads

1/ 2
2%
"
"
%
1 1
2a
a = aN $ +
1+ $ 0 ' '
$2 2
# aN & '&
#

This expression allows the calculation of the MOND gravitational acceleration provided the Newtonian one is
known.

MOND
AND GALAXY ROTATION CURVES
!
The functional form for MOND was chosen so to have a 1/r dependence of gravity at large radii. In this way flat
rotation curves are automatically explained (Figure 1). However, galaxies have rotation curve of all forms, not
necessarily flat at large radii. This is important to keep in mind. Every galaxy represents a truly independent test for
MOND, in particular when the rotation curve is not flat (Fig. 2). When fitting rotation curves note that a0 is fixed
and cannot change -- the only free parameter is the mass-to-light ratio, which is to a large extent fixed by stellar
evolution theory. Thus even though sometime one has to allow the distance of the object to vary within reasonable
values in order to achieve a decent fit, MOND gives very little flexibility. This is to be compared with the almost
unlimited adjustability of the fit offered by the dark-matter, the perfect free parameter that can appear in any quantity
and with any distribution. This fact is well exemplified in Fig 3 where an attempt is made to fit the rotation curve of
a fake galaxy. This was obtained using the photometry of one object, which defines the mass distribution, and the
rotation curve of another. Having no free parameters MOND cannot fit this curve, which is good, while dark matter
model can easily fit this non-existing object, questioning the meaning of all the fits obtained with this method.

FIGURE 1. Example rotation curve of one normal high-surface-brightness galaxy (NGC 6503, from data published in [5]).
(Left): Three-parameter dark-halo fit (solid curve). The rotation curve of the stellar (dashed line), gas (dotted line), and dark-halo
(dash-dotted line) components are shown. The fitting parameters are the M/L ratio of the disk, the halo core radius, and the halo
asymptotic velocity. (Right): The MOND fit showing the one-parameter fit (M/L). Velocities in km/s and radius in kpc.

FIGURE 2. Rotation curves for two low surface brightness, gas-dominated galaxies (data from [5]). Top: Threeparameter dark-matter halo fits (solid curve). The rotation curve of the stellar (dashed line), gas (dotted line), and darkhalo (dash-dotted line) components are also shown. The fitting parameters are the M/L ratio of the disk, the halo core
radius, and the halo asymptotic velocity. Bottom: the same rotaion curves as before, fitted with the MOND prescription.
The dotted line shows the one-parameter (M/L) fit, while the solid line show the two-parameter (M/L and distance) fit.
For UGC2259 the two lines coincide. The important thing here is that these galaxies are gas dominated so that M/L
becomes irrelevant and the fit has no free parameters at all. Velocities in km/s and distances in kpc.

FIGURE 3. Fit of the rotation curves of an hybrid galaxy, obtained using the velocity information for one
galaxy (NGC2403) and the photometry used to compute masses from another (UGC 128) (data from [15]).
MOND (left panel) fails to fit this fake galaxy, which is good. On the other hand, a fit of an isothermal dark-matter
halo (right panel) allows so much freedom that an acceptable fit can be found, which is bad. The curves relative to
the gas component (dotted line) and stellar disk (dashed line) are also shown. Velocities in km/s, and radius in kpc.

MOND AND THE T-F AND F-J RELATIONS

The Tully-Fisher relation[16] links the luminosity L of spiral galaxies to their asymptotic rotational velocity V in
the form L V4. Similarly, the Faber-Jackson[17] relation, L4, links the luminosity of elliptical galaxies the their
central velocity dispersion. Both relations find an immediate explanation within the contest of MOND because the
mass is linked to the 4th power of velocity, not the square, and in absence of dark matter mass and light are
proportional.
Indeed, equating the centripetal acceleration to the gravitational acceleration from eq. 1, and introducing the mass
to light ratio =M/L, we get:

v 4 = GMa0 "

M
L = #L
L

From this relation we see that MOND explains the T-F relation provided is roughly constant in galaxies. This is
certainly the case because all spiral galaxies share very similar stellar populations, as demonstrated by the similar
colors and spectral properties. Moreover, the fact that is basically constant allows us to introduce one important
prediction of MOND: the T-F relation is universal and all galaxies should follow the same relation.
To compare the MOND description of the T-F relation to the Newtonian one, we have to introduce the average
surface brightness of galaxies, define as the luminosity divided by the surface of the galaxy. Assuming for the sake
of simplicity that galaxies are round, we have =L/r2. Then equating the centripetal acceleration to the acceleration
of gravity we have:

v4 =

(GM) 2 M 2 #
" 2 L = $ 2 #L
2
r
L
2

Here we can see the T-F relation provided the product " # is constant. As we just pointed out is basically
constant, thus we have to conclude that also is constant. This is certainly not the case because galaxies divide in
two big families, the high surface brightness (HSB) and low surface brightness (LSB) galaxies. The latter having up
to 100 times fainter surface brightness than normal HSB galaxies. Newtonian dynamics therefore predicts HSB and
LSB galaxies should follow two parallel but distinct
! T-F relations. This is simply not the case (Figure 4). Both HSB
and LSB galaxies follow the same T-F relation, as predicted by MOND. When Milgrom first made these
considerations, no dynamical information were available for LSB galaxies, so this was a true prediction.

FIGURE 4. (Left) The Tully-Fisher relation for a sample of low surface brightness galaxies (points) is compared to the
relation defined by HSB galaxies (solid line). Data refers to the two homogeneous datasets presented in [18]. As
predicted by MOND and contrary to the Newtonian prediction, HSB and LSB galaxies follow the same relation. (Right)
The Faber Jackson relation for elliptical galaxies links the central velocity dispersion to the total luminosity. The line
correspond to L/L = 104 (with in km/s). Reproduced from [17].

MOND AND ELLIPTICAL GALAXIES

Due to the high central density a considerable fraction of elliptical galaxies is in Newtonian regime of
accelerations. This is why the properties of elliptical galaxies can be explained to a large extent without the needs for
dark matter. This also means that in first approximation the virial theorem applies. Velocity dispersion , size, and
mass of elliptical galaxies are therefore linked and when represented in the space defined by these three variables,
they should define a plane. Mass, however, is never directly observed so that to verify whether this is true, one has to
find a way to estimate it. Within the contest of MOND this is trivial, all we need is to convert the observed light into
a mass via , which is sensible constant among ellipticals. Alas, MOND predicts that when represented in the space
defined by luminosity, size, and velocity dispersion all elliptical galaxies should follow a plane (a surface to be
precise, as we will se in a moment).
By contrast, if we believe galaxies are filled by huge amounts of dark matter, distributed in arbitrary ways, there is
no relation between luminosity and mass. The definition of size also become fuzzy, because dark matter is much
more extended than the visible galaxy indeed the end of dark haloes was never observed so far -- that one can
not use the observed radius in the virial formula. As a consequence, elliptical galaxies should not follow a plane in
this particular space, something in open contrast with the fact that elliptical galaxies are known to define a plane -the fundamental plane-- in full agreement with MOND prediction.
To explain the existence of the fundamental plane (FP) one has to assume that dark and luminous matter are mixed
in a constant ratio in elliptical galaxies. Even if this were true, however, things are not fine because the fundamental
plane is tilted with respect to the plane defined by the virial relation (Fig. 5).
Finding an explanation for the tilt has been and still is one of the major puzzles in modern astrophysics. In a nutshell,
both possibilities of a systematic change of the mass-to-light ratio with luminosity and departure from homology
were fully investigated. To preserve the narrowness of the FP, the tilt requires dark matter to be finely tuned to the
galaxy luminosity, or the departures from homology to be finely tuned to the galaxy size[19]. Of the two options the
former is the most widely accepted, and the tilt ascribed to L0.25 (e.g., [20]). Whatever the reason for the tilt is,
both these possibilities face paramount difficulties as soon as LSB elliptical galaxies are considered, because these
galaxies define a completely different plane[21,22] implying L-0.4.
As noted before, in absence of dark matter Newtonian dynamics predicts galaxies to follow a perfect plane. This is
only approximately true for MOND. The central region of elliptical galaxies is in Newtonian regime while the
outskirt is not. Thus, according to the fraction of the galaxy that is in MOND regime remember the effective radius
used to place galaxies in the plane contains always half the total light -- galaxies deviate more or less significantly
from the virial relation. How important MOND effects are on galaxies of different size/type is determined by their
average surface brightness (Fig. 6). This is seen to vary in a systematic way. HSB galaxies became progressively
less dense as they increase in size or luminosity. For them the MOND effects are therefore increasingly relevant and
deviation from the virial relation becomes progressively more evident as we move from small to large galaxies. On
the other hand, it is seen that LSB galaxies become progressively more dense as their size increase, and MOND
effect get smaller. This is exactly what is observed, and the trend defined by MOND fully agrees with the measure
slopes (Figure 7).

FIGURE 5. Edge on view of the fundamental plane of elliptical galaxies. The quantities k1 and k3 as defined in [19] are used. In
Newtonian dynamics, k1 is proportional to the total galaxy mass, while k3 is proportional to . The two dotted lines with slope 0.4 and +0.25, show the dependence of on M for LSB and HSB, respectively. Data were taken from [19]: Squares= giant
ellipticals; Triangles= intermediate ellipticals; Diamonds= dwarf ellipticals; circle= compact ellipticals; Crosses=bulges; Dots=
LSB ellipticals; Stars= LSB elliptical from [22]; Pentagons= dwarf elliptical in Virgo from [23]; Pluses= radio galaxies from
[24].

FIGURE 6. (Left): The Kormendy relation showing the trends of the mean surface brightness as a function of the effective
radius for HSB and LSB elliptical galaxies. HSB galaxies becomes progressively more diffuse with size, while LSB galaxies do
exactly the opposite, becoming more dense as their size increase. (Right): The trends in surface brightness are just trends in
internal acceleration of gravity aN. The different trends of the mass-to-light ratio observed in HSB and LSB galaxies can be
understood in terms of the different strength of the MOND effects. Symbols as in Figure 5.

MOND AND ULTRA COMPACT DWARF GALAXIES

Dwarf galaxies have attracted the attention of astrophysicists in the last years for many reasons. Among others,
these galaxies are found to contain an impressive amount of non-baryonic dark matter[25,26] having mass-to-light
ratios up to 100. The discovery[27,28,29] in an all-object survey of the Fornax cluster of a new type of galaxies, the
ultra-compact dwarfs (UCD), has added a new member to the dwarf family. Unlike other dwarfs, however, UCD
galaxies have very high central concentration and thus star-like morphology in typical one-arcsecond-resolution
ground-based imaging. At the distance of the Fornax cluster (20 Mpc; Hubble constant H0 = 75 km s-1 Mpc-1) this
implies sizes of 100 pc at most. The B band luminosity ranges from to 3107 L, significantly brighter than the
brightest globular cluster known. Thus, UCDs are squarely placed midway between regular dwarf galaxies and
globular clusters.

Figure 7. Plot of a/aN = /0 versus luminosity (symbols as in Figure 5). The meaning of this plot can be understood noticing
that the discrepancy in acceleration is interpreted as due to extra mass or, in other words, as a variation of the mass-to-light ratio
with respect to the true value 0. The mass discrepancy decrease (increase) with luminosity for LSB (HSB) galaxies. The
dependence is indicated by two dotted lines, with slope L-0.3 for LSB and L0.25 for HSB galaxies, respectively. In absence
of dark matter k1 L, thus this plot is equivalent to Fig 5. The observed slopes for a/aN = /0 are teh explanation of the tilt in the
framework of MOND. The dashed and solid lines are the same drawn in Fig. 6 and are meant to visualize the loci followed by
LSB and HSB galaxies.

Recent determination[30] of their effective radius (ranging from 10 to 22 pc) and velocity dispersion (ranging
from 24 to 37 km s-1) indicates that, if any, there is very little dark matter in UCD galaxies. Indeed assuming UCDs
are virialized structure, masses ranging from 107 to 108 M are found[30]. The corresponding varies from 2 to 4 in
solar units, fully consistent with the expectation for an old stellar population without significant amount of dark
matter. In the literature two different scenarios have been proposed to explain why UCDs show no mass
discrepancy. One possibility is that UCDs are giant globular clusters[31] possibly the result of the merging of the
giant stellar clusters created during periods of strong galaxy interaction[32], like the one observed in the ``Antennae''
system (e.g., [33]). The other possibility is suggested by the fact that UCDs luminosity and size match well the one
of the nucleus of dwarf galaxies. It is therefore possible that UCD galaxies are the remnant of normal nucleated
dwarf galaxies that have lost their external halo and dark matter, because of strong and repeated interaction with
other galaxies. This latter scenario, referred to as galaxy threshing[34], is the most widely accepted.
Thinking in terms of MOND, the word compact is sufficient to know that no mass discrepancy should be found
in these objects. To show that is the case, lets put a lower limit to the internal acceleration of gravity taking the less
luminous UCD having luminosity 4106 L, assigning to it the maximum possible diameter 100 pc, and the
minimum mass-to-light ration (=2). At the outer edge of this galaxy the acceleration is 210-8 cm s-2 > a0.
According to MOND, then, UCDs are everywhere in Newtonian regime and no non-baryonic dark matter is to be
found. Up to now, MOND is in full agreement with observations.
It is also clear that the reason why we do not observe a mass discrepancy is neither due to the particular
evolutionary history of this galaxies, nor to the fact that UCDs might be giant globular clusters. It is solely due to the
fact that they are in Newtonian regime and therefore should strictly obey Newtonian dynamics.

MOND AND GALAXY CLUSTERS

Moving to cluster of galaxies we see that also for these structures MOND gives reasonable result[10]. However, rich
galaxy clusters do represent the only objects where the agreement is not as good as we have seen so far. It is evident
from Figure 8 that the mass discrepancy does not completely disappear, there is a remaining systematic discrepancy
between the MOND dynamical mass and the observed mass. How severe this is for MOND is not clear. Certainly it is
at least a good sign that MOND masses are larger that the total mass observed, indicating some baryonic component
of the clusters still remains undetected. If the discrepancy were the other way around, it would be devastating for
MOND.
In many astrophysical contexts, a factor of 2 discrepancy can often be accommodated by reconsidering the effects of
the several idealized assumptions that are not realized in every case. A fair number of simplifying assumptions is

certainly necessary in order to be able to derive cluster masses, so it would not be surprising if this discrepancy will
disappear in the future.
However it is quite possible, and perhaps even good for MOND, that the residual mass discrepancy is real and that
the accounting of baryons in clusters is not yet complete. Indeed, this seems to be required, even in the context of
MOND, by strong gravitational lensing.
The critical surface density required for strong lensing is

"c =

1 cH 0
F
4# G

where F~10 is a dimensionless function of the lens and source redshifts[35], c is the speed of light, and H0 the
Hubble constant. MOND applies at surface densities[2] below ~ a0/G ~ c/5, that is to say, the critical surface
density for strong lensing is always greater than the upper limit for MOND phenomenology. Strong lensing never
occurs in the MOND regime. Strong lensing observed in clusters typically requires a total projected mass in the inner
100200 kpc between 1013 and 1014 M, which is not present in the form of hot gas or stars. Whether or not MOND
is correct, undetected matter does seem to be required in the cores of rich clusters. Interestingly, masses of this order
of magnitude would make the residual discrepancy disappear. Thus one can imagine the MOND value of the
dynamical mass is correct, and a prediction is made that the remaining baryonic mass to be found in cluster is within
a factor 2 of the currently observed mass.

FIGURE 8. (Left) The Newtonian dynamical mass for galaxy clusters plotted against the total observable mass
(gas+stars). The solid line represents the locus of no mass discrepancy. (Right) The factor ten discrepancy is reduced
by MOND down to a factor two. Reproduced from [10].

MOND AND GRAVITATIONAL LENSING

As seen in the previous paragraph a strong gravitational field is needed to significantly deflect light and produce
multi-images of a background source. Thus strong gravitational lensing always occurs in Newtonian regime and do
not constitute a test for MOND.
Weak lensing, however, do occur in the weak field regime and we can use the weak tangential alignment, the
cosmic shear, of distant galaxies caused by nearby ones to test MOND.
To do the test we need to compute the effects of MOND on light. In general relativity a photon experiences twice
the deflection of a massive particle moving at the speed of light. Thus, as normally done in Newtonian dynamics, in
first approximation in the contest of MOND one has to compute the deviation for a massive particle and than double
the effect. This is what was done so far to test MOND.
Considering the simplest case of a test particle that does not disturb the external mass distribution, and with
speed such that its path is nearly linear, the deflection can be approximated by integrating the acceleration along the
unperturbed trajectory (i.e. a straight line). Further restricting to the case of a point mass M, the asymptotic
deviation angle (in radians) is

AM =

2"
GMa0 .
c2

The deflection of light at large impact parameters from a point mass in the MOND framework is independent
from the impact parameter, as much as the rotational velocity is independent from distance. This can be easily
understood because the MOND acceleration
at large impact parameters is proportional to 1/r, while the period of
!
time over which this acceleration applies is proportional to r. In this way the deflection of light at increasingly large
impact parameters is constant (Figure 9).
The deflection angle is not directly measurable, but the distortion of images the shear -- and the (total)
magnification of sources can be calculated directly from A M , and these are observable.
So far the cosmic shear has been used to derive the dark matter content of the deflectors (e.g. [36]). Not
surprisingly, results are all consistent with galaxies being surrounded by very extended and massive isothermal
haloes. Interestingly, no upper limit has been placed on the halo size, despite the fact that a systematic distortion has
been measured out to hundreds of times the visible extent of the foreground galaxies. This is by far the most
stringent lower limit on the size of dark matter haloes that, to all practical effects, seems to be infinite.
MOND results are once again consistent with observations (Figure 10).

FIGURE 9. The deflection of light at large impact parameters D is constant because a 1/r while the time over
which this acceleration applies is dt=rd/cr. The net effect is that the deviation of light depends only on the mass
M of the object.

FIGURE 10. The mean cosmic shear around foreground galaxies compared to Newtonian and MOND
prediction. The original data [37] in the g, r, and i photometric bands have been averaged, excluding the first and
third g points that clearly deviated from the values found in the other two bands The models shown are the one
presented in [38] and refers to the MOND prediction (solid line), and the Newtonian result if there is no dark matter
(dashed line).

MOND AND THE BACK HOLE BULGE MASSES RELATION

As soon as the mass (MBH) of a number of black hole located at the center of galaxies were measured it was
recognized that MBH is related to the global properties of the galaxy. First a correlation was noted[39] between MBH
and the luminosity of the spheroidal component -- the bulge of the form MBH/M=0.01LB/L (Figure 11). Later a
possibly tighter relation MBH/M=0.074 was found between MBH and velocity dispersion of the bulge[40, 41]. The
physical processes responsible for such correlations remain unclear, though they are important because they help
understanding the BH-galaxy formation process, also offering a tool to estimate BH masses by the much easier
direct observation of the global properties of the host galaxy. The problem here is that in the presence of dark matter
it is difficult to reconcile the two relations. While is somehow proportional to the bulge mass (including dark
matter), the bulge luminosity LBulge is certainly not. Nevertheless, the two relations are necessarily linked because
both are pointing to the same thing, the BH mass. Because of this inconsistency, the two major groups involved in
the debate soon started arguing whether only the MBH- relation should be considered and the MBH-LBulge, due to the
somewhat larger scatter, neglected. A war also ignited around the correct slopes for both relations and up to now
no agreement was reached.
By contrast, according to MOND Nature has chosen the simplest solution (always preferable): every bulge puts a
fix fraction of its total mass into its central BH.
Indeed, in absence of dark matter, we have MBH/M=0.01LB/L =0.005MBulge/M (for =2). Similarly, keeping in
mind the F-J relation (Figure 4) we have MBH/M=0.074 = 0.007LBulge /L= 0.0035MBulge/M fully consistent with
the previous one.
MOND here is relevant in telling us that the MBH-LBulge has slope exactly 1, and that the right exponent for the
MBH- relation is exactly 4, value that just happens to be the average of the two values proposed by the two major
groups involved in this field.
Thinking out of the box, one can point out that MOND implies the F-J relation to be universal for spheroids,
meaning also LSB galaxies should follow the same relation for BH mass. At present it is not at all clear whether
such diffuse objects can form a central massive object so it will be interesting to make experiments to see whether
this is the case and, if yes, how BH masses compare with this prediction.

FIGURE 11. (Left) The black hole mass bulge central velocity dispersion relation is well described by the
relation MBH/M=0.074 (solid line, in km/s). (Right) The bulge luminosity black hole mass relation is
consistent with the linear relation MBH/M=0.01LB/L. Both relations are well understood in absence of dark matter.
Data are from the compilation of [42].

MOND AND GLOBULAR CLUSTERS

Of the many possible tests for MOND the most effective are the one involving the study of objects believed to be
free from the effects of dark matter. The only test of this type so far attempted involves globular clusters. It is well
know that the properties of these objects can be explained without resorting to dark matter. Thus these objects should
follow Newtonian dynamics down to whatever weak acceleration, unless Newtons law breaks down below some
acceleration.
In the framework of MOND globular clusters agree to Newtonian dynamics because they are extremely compact
and the internal gravitational field strong. Moving far enough from their center, however, MOND predicts the field
will eventually deviate from the Newtonian one. If this is the case, then it will be difficult to explain the deviations
specifically because dark matter cannot be invoked. A first pilot experiment for the globular cluster Omega Centauri
was presented in [43] and a full account of the status of these experiments is given elsewhere in this book (Scarpa,
Marconi & Gilmozzi). Here, it should suffice to say that globular clusters do show MOND effects as soon as their
internal gravitational field goes below a0, fully confirming the MOND phenomenology also in these class of objects
and perhaps representing the ultimate test against the existence of non-baryonic dark matter.

CONCLUSIONS
It should be clear by now that wherever the acceleration of gravity goes below a0 we should detect MOND effects.
This general statement, combined with a very simple prescription for computing gravitational accelerations in the
weak filed limit, is extremely effective in describing the properties of many different objects seen in the universe.
For instance, lets have a look at Figure 12 where dynamical and luminous masses are compared for about 1000
objects. In this figure we see how precisely MOND eliminates the mass discrepancy in objects ranging from
globular clusters to clusters of galaxies that are at least 104 times bigger and 107 times more massive. Not only, it
also explains why objects as diverse as LSB dwarf galaxies and group of galaxies do show the same mass
discrepancy inspite of the vastly different physical properties. A particularly interesting case is the one of the
globular cluster Palomar 13 that has been reported to have unusually large velocity dispersion. This is a very diffuse
cluster with internal accelerations of the order of 10-9 cm s-2 that looks like an open cluster but is as old as globular
clusters. Well, Palomar 13 do follow the MOND law and show the same mass discrepancy of galaxies of
comparable surface brightness, as is clearly seen in Fig. 12.
Looking at this figure, one has the feeling that MOND applies universally. Milgrom, however, has repeatedly stated
the MOND formalism applies only to objects for which the total gravitational field in below a0. That is to say, in
determining whether MOND effects are relevant for a given object, any external contribution from nearby objects
must be taken into account. The reasons for this claim rest on the fact that nearby open stellar clusters which
experience a field of the order of a0 due to the Milky Way while having a very weak internal field -- do not show
large mass discrepancies. To explain this apparent inconsistency, Milgrom was led to conclude that what is relevant
is the total field. All tests but one done so far do satisfy this restriction (total field below a0). We therefore are not in
the position to tell whether MOND applies only when the total acceleration is below a0 or if the breakdown of the
Newtonian dynamics occurs irrespectively of the external field. According to Milgroms point of view, MOND
cannot be tested within the solar system and one as to move as far as 0.1 light years from the sun before MOND
effects could be detected. Thus, depending on the proximity to massive objects, an object normally showing large
mass discrepancies should follow the Newtonian dynamics below a0. A good example of this situation is the case of
globular clusters orbiting nearby the Milky Way. Up to now MOND was tested in three globular clusters one of
which, NGC 6171, was studied precisely for being close enough to the Galaxy that the total field determining the
dynamics of its external regions is above a0. Nevertheless, this cluster seems to follow the MOND law. At the time
of writing this result, being limited to one object, is not conclusive so we are not in the position of drawing any firm
conclusion. More data are being collected at the European Southern Observatory VLT telescope in order to clarify
the issue.
It is worth to point out that open clusters are not in dynamical equilibrium and are known to disrupt rather quickly.
For instance, the vast majority of open clusters are not able to survive even for a single complete rotation of their
own host galaxy. Their internal velocity dispersion, therefore, is not representative of their total mass and some other
effect, yet to be understood, are responsible for it. In my opinion, open clusters put no constraints on MOND and the
claim by Milgrom is unjustified.
Thus, it is my personal opinion and I am the only one responsible for it if proved wrong-- that if Newtonian
dynamics fails below a0, this should be true irrespectively of the total field and one should be able to observe MOND
effect also here on earth. For instance, I think a refined version of the Cavendish experiment studying gravitational
forces in the horizontal plane should detect MOND effects.

Figure 12. Comparison of the Newtonian acceleration of gravity computed using dynamical masses, versus luminous masses
(i.e., converting light into mass via ) for about 1000 objects. Symbols as in Figure 5 (Left) In the upper left corner are globular
clusters (triangles) for which no mass discrepancy is observed. Moving to galaxies (squares and crosses) we start seeing deviation
that become evident when considering LSB galaxies. Groups of galaxies (pluses) do show an enormous mass discrepancy. Rich
clusters (circles with pluses) are the only class of objects that seem to deviate systematically from MOND prediction. Finally, two
large pentagons represent the globular cluster Omega Centauri and Palomar 13. The solid line shows the locus of objects for
which dynamical and luminous masses agree while the MOND value of the acceleration -- computed from luminous masses is
shown by the dashed line. (Right) Due to the inevitable dispersion of the data, the agreement between MOND and observations
became more evident after rebinning. It is seen that MOND agrees with observations over at least 6 orders of magnitude in
acceleration.

ACKNOWLEDGMENTS
I am grateful to the dr. Eric Lerner and dr. Jose Almeida for making this conference possible and for giving me
the possibility to freely express my idea, something that would have not be possible in other forums.

REFERENCES
1. Binney J. 2004, in Dark Matter in Galaxies, ASP conference series, IAUS 220, 3
2. Milgrom M. 1983a ApJ 270, 365
3 ------ 1983b, ApJ 270, 371
4 ------ 1983c, ApJ 270, 384
5 Begeman K. G., Broeils A.H. ans Sanders R. H. 1991, MNRAS 249, 523
6 Milgrom M.1994, ApJ 429, 540
7 Sanders R. H. 1996, ApJ 473, 117
8 McGaugh, S. S. and de Blok, W. J. G. 1998, ApJ 499, 66
9 Milgrom M. 1998, ApJ 496, L89
10 Sanders R.H. 1999, ApJ 512, L23
11 Qin B. and Zou Z. L. 1995, A\&A 296, 264
12 Mortlock D. J. and Turner E. 2001, MNRAS 327, 557
13 Sanders R.H. and McGaugh S.S. 2002, Annual Rev. Astron. Astrphys. 40, 263
14 Milgrom M. 1984, ApJ 287, 571
15 de Blok W.J.G. and McGaugh S.S. 1998 ApJ 508,132
16 Tully, R.B.; Fisher, J.R 1977, A&A 54, 661
17 Faber S.M. and Jackson R.E. 1976 ApJ 204, 668.
18 Zwaan M.A., van der Hulst J.M., de Blok W.J.G, and McGaugh S.S. 1995, MNRAS 273, L35

19 Ciotti L., Lanzoni B. and Renzini A. 1996, MNRAS 282, 1

20 Bender R., Burstein D. and Faber S. M. 1993, ApJ 411, 153
21 Dekel A. and Silk J. 1986, ApJ 303, 39
22 Peterson R. C. and Caldwell N. 1993 AJ 105, 1411
23 Geha. M., Guhathakurta P., and van der Marel R.P. 2002 AJ 124, 3073
24 Bettoni D., Falomo R., Fasano, G., Govoni F., Salvo M. and Scarpa, R. 2001 A\&A 380, 471
25 Mateo M. L. 1998 ARA\&A 36, 435
26 Kleyna J.T., Wilkinson M. I., Evans N. W., and Gilmore G. 2001 ApJL 563, 115
27 Hilker M., Infante L., Vieira G., Kissler-Patig M., and
Richtler T. 1999 A\&AS 134, 75
28 Drinkwater, M.J., Jones J.B., Gregg M.D. and Phillipps S. 2000, PASA 17, 227
29 Pillipps S., Drinkwater M.J., Gregg M.D. and Jones J.B., 2001 ApJ 560, 201
30 Drinkwater, M.J., Gregg M.D., Hilker M., Couch W.J., Ferguson H.C., Jones J.B., and Phillipps S. 2003, Nature
423, 519
31 Mieske S., Hilker M., and Infante L. 2002 A\&A 383, 823
32 Fellhauer M. and Kroupa P. 2002, MNRAS 330, 642
33 Whitmore B.C., Zhang Q., Leitherer C., and Fall S.M. 1999, AJ, 118, 1551
34 Gregg M.D., Drinkwater M.J., Hilker M.J., Phillipps S., Jones J.B., and Ferguson H.C. 2003 Ap\&SS 285, 113
35 Blandford R. D. and Narayan R. ARA&A 30, 311
36 Brainerd T.G., Blandford R.D., and I. Smail 1996 ApJ 466, 623
37 Fisher P. et al. 2000, AJ 120, 1198
38 Mortlock D.J. and Turner E.L. 2001 MNRAS 327, 552.
39 Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115,
40 Ferrarese, L. \& Merritt, D. 2000, ApJ, 539, L9
41 Gebhardt, K., Bender, R., Bower, G., et al. 2000, ApJ, 539, L13
42 Tremaine, S., Gebhardt, K., Bender, R., et al. 2002, ApJ, 574, 740
43 Scarpa, R., Marconi, G., \& Gilmozzi R. 2003A, A\&AL 405, 15